Teaching Responsible AI (TRAI)

Currently, many organizations are adopting artificial intelligence (AI), and want their organizations to become more AI-driven. Most of them are already taking steps towards increasing their AI maturity. However, applying AI technology requires more than technological expertise. 

By now, companies and public institutions that have developed their AI maturity to substantial levels will mostly harness an abundance of technical experience. But at the same time, their technical workforce often lacks the required domain knowledge and experience to understand how the business and other stakeholders could either benefit from or risk being harmed by AI-driven innovation. In order to apply AI technologies in an effective and responsible manner, organizations must develop a solid understanding of the intended and established (ethical, legal, societal, and environmental) impact of the solutions they are implementing within their organization as well as within their broader field of work. And so the people working in technical departments must develop knowledge and skills needed to understand and mitigate the particular risks of applied AI within their company’s specific sector (e.g. unfairness and unequal treatment, opaqueness / black boxes, unreproducible results, etc.). 

Ambition

The current working group aims to aid students and professionals within the field of AI to learn more about everything related to the responsible (and sensible) application of AI within a company or organization. 

On the one hand, reflection on the broad impact of AI and on its ethical, legal and societal aspects should be an integral part of any Artificial Intelligence or Data Science curriculum or module. As any decision in a design or development process has ethical consequences and is subject to legal boundaries, these aspects should ideally not be taught in separate courses, but rather run throughout the curriculum. As suggested by (Miller, 1988), technical issues are best understood in their social context, and the societal aspects of computing are best understood in the context of the underlying technical realization.

To establish true impact, we aim to bring together educational institutes and the organizations where students end up working after graduation. By involving both the educational institutes and those organizations, we can ensure that the educational programs teach students the skills needed to help innovate the companies and public institutions of tomorrow, and that these organizations are ready to embrace and make good use of the knowledge brought in by the newly graduated.

The goal of the TRAI working group is to bring together TRAI stakeholders who are interested in developing materials, practices, and methods of teaching and adopting Responsible AI.  Stakeholders can for example be teachers and students of higher education (current focus is on HBO/WO, we aim to expand to include MBO as well) in Computer Science, AI and related disciplines; upper management, employees, and talent development professionals of Dutch institutions and companies that use AI systems and develop AI policies and strategy; as well as other beneficiaries. We use the term Responsible AI as a broad term referring to the importance of accounting for ethical, legal, societal, environmental and human-centered aspects when developing AI systems.

Aims

Bringing these groups together will facilitate achieving the following specific aims:

Exchanging expertise

Teaching and adopting Responsible AI is challenging due to the inherent multidisciplinary nature of the subject. Responsible AI requires on the one hand deep technical knowledge about AI methods and techniques, and on the other hand insights into the way AI is and could be embedded into organizations and society and associated ethical, legal and societal consequences and risks, combined with an understanding of how one affects the other. Through this working group we bring together teaching practitioners, students and industry and institutional stakeholders of different fields to allow for sharing of expertise, methods, pitfalls, best practices and practical cases in teaching and adopting Responsible AI. In this way we will be able to accelerate the integration of responsible aspects of AI into our curricula and engineering practices, and professionalize our methods and approaches for teaching and adopting Responsible AI.

Community building

Besides exchanging expertise on the content and methods of teaching and adopting Responsible AI, we furthermore aim to create a community through which one can meet others with an interest in the topic and share experiences. The community can also provide support and help for those new to teaching Responsible AI in getting started and finding the necessary resources, thereby lowering the threshold of entering the area. 

Agenda setting

Although the importance of Responsible AI is more and more recognized, the traditional separation of engineering and social sciences still informs much of our current teaching and engineering practices. By coming together through this working group, we will increase our impact in putting this topic on the agenda of educational institutions and industry. It will allow us to create a better connection between what is taught and what is practised by, on the one hand, bringing in the visions, hesitations, experiences, obstacles, and so on from engineering practices into our educational programs, while on the other hand sharing the latest scientific insights with industry. Finally, by bringing together a TRAI community we will be able to identify gaps and opportunities across organizations and sectors, e.g. for creating shared educational modules.

TRAI inventory

The first TRAI activity that we are setting up is an exploratory inventory to get a better understanding of how and to what extent Responsible AI is addressed in current AI and Data Science educational programmes in higher education, and which opportunities and challenges arise in shaping this. Moreover, we aim to understand which knowledge, skills, and attitudes are needed within (AI and data science departments of) Dutch companies, institutions, and organizations regarding the implementation and responsible use of AI.

We are planning to interview several AI teachers and directors of educational programmes on AI and Data Science, as well as stakeholders within public institutions and companies. We will discuss questions such as where Ethical, Legal and Societal Aspects (ELSA) are taught in curricula, which types of topics are taught, and which challenges arise in setting up education around these topics.

We will produce a report (expected late 2023) which provides an overview of the most important insights and needs across institutions and organizations regarding education of Responsible AI. 

Community building

The TRAI inventory is intended to form a starting point for shaping further TRAI activities regarding exchange of expertise, community building, and agenda setting. 

We anticipate the organization of a first community event sometime in 2023. If you would like to receive updates about our upcoming activities, please contact us.